Nowadays, among the most important communication channels are (i) the Internet, and (ii) smartphones, which of course can be also used as a gateway to the Internet. Globally, there is a growing debate as to whether an overuse of mobile phones and the Internet can be defined as a form of a behavioral addiction [1
]. Overuse clearly cannot be understood simply by referring to the time spent on mobile phones or on the Internet, but rather other addiction-related concepts such as preoccupation, withdrawal, development of tolerance or personal suffering because of the usage are more important variables. The importance of the topic is reflected by the large number of conducted studies [4
] and also by the inclusion of Internet Gaming Disorder as an emerging disorder in Section III of the DSM-5. More and more focus is also on smartphone addiction [6
]. Smartphones represent a more sophisticated version of the overarching category of mobile phone. Given the controversy about the nature of smartphone overuse, “problematic mobile phone use” might be a better term compared to “mobile phone addiction” or “smartphone addiction”. However, given the handiness of the term “addiction”, we use both the terms “problematic use” and “addiction” synonymously in the present manuscript. A further note: Although the terms “mobile phone” and “smartphone” are not exactly the same (as outlined above), for reasons of simplicity and also because of the fact that we study classic variables such as call and SMS behavior, the terms are used somewhat interchangeably in the present study. It is worth noting that we did not study usage of social messenger channels such as Facebook or WhatsApp in the present research endeavor.
In a recent paper, we outlined the need to include methods from computer science in diagnostics to get a fuller picture of psychopathological disorders [9
]. In that paper, we also coined the term “psychoinformatics” describing the administration of computer science methods (such as tracking smartphone behavior) to study psychological phenotypes. The same term has also been introduced by Yarkoni [10
]. Not only could diagnostics of behavioral addictions being related to technology (e.g., excessive online video gaming, online gambling, mobile phone usage) benefit from the inclusion of actual recorded behavior, but so too could the course of therapy of these non-substance related addictions [11
]. Additionally, mobile phone applications can also be of value in the treatment (e.g., aftercare) of addictions such as alcoholism [12
Naturally, actual behavior should constitute a better predictor for addictive tendencies than self-reported variables. As attractive as this idea seems, only some recent studies could in parts back up this idea [13
]. The present study aims to find further support, while investigating excessive mobile phone behavior (with a focus on smartphones). We compared self-reported data with directly recorded data over several mobile phone variables (including calls and SMS) to predict mobile phone addiction. It has earlier been shown that recorded behavior differs from self-reported data on mobile phone usage [13
]. Beyond this research question, we asked ourselves if actual recorded mobile phone behavior in hours is more closely related to self-reported addictive mobile phone usage tendencies compared to self-assessed mobile phone behavior in hours.
First, we hypothesized that participants would have problems assessing their mobile phone usage, thereby (according to [14
]) underestimating their actual use. Second, we hypothesized that actual recorded mobile phone behavior compared to the self-reported assessments should be more closely associated with results from the mobile phone addiction questionnaire.
The present study compared relevant variables for problematic mobile phone usage, as recorded directly on the phone, with self-reported assessments from the users. To some degree in line with our first hypothesis, we observed that the aggregated length of weekly mobile phone usage in hours was overestimated by the participants of the study, while more distinct behaviors, such as weekly outgoing calls, have been underestimated. Of note, the overestimation effect would have been significant if results from a one-sided test procedure were presented (or arguably with a larger sample size). These summarized results indicate that the participants were not able to estimate their mobile phone usage in correct numbers. Our study results differ in parts from those presented by Lin et al.
], who only reported underestimation of smartphone usage in their study. This could be due to differences in questions asked of participants. According to Lin et al.
(2015), they asked “participants how many hours [they spent] on their smartphone on average during a weekday, and then asked if there was any difference between their weekday and weekend use” [14
] (p. 141). In contrast we asked our participants simply how much they estimated their weekly usage to be in hours. Furthermore, we also addressed more specific activities such as incoming calls. Moreover, awareness of the topic of study might have led to an exaggerated number of total mobile phone usage in our study. However, this might also be true for the Lin et al.
study. Finally, in the Lin et al.
study a substantial part of the sample was characterized as being smartphone addicted (31 out of 79 participants), whereas our sample seemed to be largely in the normal usage range. For further illustration: Mean of the MPPUS was 54.47 (SD = 14.45); median = 51.50 (IR = 20.75)—please remember that scores range between 27 and 135 (our data set ranged from 30–89).
In line with our second hypothesis, the overall pattern of correlations between the recorded and self-reported variables and the mobile phone addiction scores demonstrates that the recorded behavior is more strongly associated with addictive tendencies—two out of five associations would have not been found when only asking the participants. These findings illustrate the potential benefits from close collaborations between computer science and psychiatry/psychology, which would allow the inclusion of direct tracking of behavior. These methods could aid the diagnostic process and therapy by objectively recording the behavioral addiction of interest [11
We are convinced that the present results can be transferred to some extent to other technological behavioral addictions such as Internet addiction. We come to this conclusion because smartphone addiction and Internet addiction overlap as outlined by Kwon [6
]. Moreover, both forms of addiction rely on technology use, which can be recorded. As outlined by others [14
], users of electronic devices seem to have problems assessing the time they spend online. Psychoinformatics will help them to get exact feedback on the actual media consumption. How psychoinformatics will be of relevance for diagnostics and treatment of substance-related addictions is an important question for further research. Here, text mining, the statistical analyses of text in SMS and emails, can reveal information about the mood of a person (by observing and counting the use of negative and positive words). Although this might be a great leap from the present results, insights into emotional states are of high importance when studying addiction (e.g., withdrawal goes along with negative emotionality). Other functions such as GPS tracking will be able to give new insights into addiction such as that a very stressful life in terms of excessive mobility might be associated with other forms of addiction, even substance related addictions like higher alcohol consumption. (e.g., Internet addiction has repeatedly been associated with higher alcohol consumption) [16
At this point, we want to discuss a number of limitations of the present study. In the current study, we decided to look at classical smartphone variables exclusively (primarily to avoid the problem of multiple tests when exploring numerous variables). However, a future study will need to monitor activities in social networks or other instant messengers such as WhatsApp, which are currently heavily consumed by users and might be even more strongly associated with mobile phone and smartphone addiction compared to the present variables of interest. In this context, we refer to a new study [18
] showing that WhatsApp usage is one of the driving forces of smartphone usage. Additionally, the present study investigated a rather small number of participants stemming from university campuses (thus, making it hard to take gender and age into account as control variables). As well, the investigation of clinical samples in the context of addictive or problematic behavior clearly is also warranted. Consequently, the results need to be replicated in more representative samples or participants at stronger risk for the problem behavior under investigation. Other variables such as number of logins on the smartphone (and hence the fragmentation of everyday life) might be also very interesting. Finally, we discarded the first week of data in order to tackle the problem that the recorded behavior of the present study might have been influenced by feelings of being observed. This approach was based on no empirical evidence and future studies need to provide numbers to deal with this problem more adequately.
Nevertheless we are of the opinion that the empirical data of this study represent an important (although preliminary) starting point to illustrate the advantages of including computer science methods in psychology and psychiatry.
However, even after taking into account some of the advantages of objectively recorded data, one can imagine that psychology and psychiatry would not be possible without self-report data. In our opinion, self-report data was and will always be an important data source when dealing with well-being or the mental state of a person. In this context, we hold that a combination of self-report data and actual recording of the problem behavior will provide the clearest picture of the patient, as well as giving insight into to the ability of the patient to reflect about his or her addiction. This is underlined by some data in Table 1
: the sometimes differing means/standard deviations (either high or low) show that self-assessment and actual behavior clearly are not the same.